rjac/whisper-medium-BTCv2

Whisper-medium-BTC

This model is a fine-tuned version of openai/whisper-medium.en on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3875
• Wer: 6.5315

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 3e-06
• train_batch_size: 32
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 600
• training_steps: 800
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.7348	2.01	50	0.7378	9.3091
0.5276	4.02	100	0.5290	8.6125
0.3585	7.0	150	0.3875	6.5315
0.2924	9.01	200	0.3548	6.6779
0.2506	11.02	250	0.3364	6.7888
0.1946	14.01	300	0.3262	7.1482
0.1411	16.02	350	0.3329	7.2104
0.1005	19.0	400	0.3422	7.5565
0.0535	21.01	450	0.3532	7.1793
0.0259	23.02	500	0.3456	7.5121
0.0137	26.0	550	0.3587	7.6541
0.0078	28.02	600	0.3591	7.3524
0.0041	30.02	650	0.3672	7.3035
0.0026	33.01	700	0.3962	7.3213
0.0022	35.02	750	0.3997	7.3524
0.0022	38.0	800	0.4025	7.3302

Framework versions

• Transformers 4.27.0.dev0
• Pytorch 1.13.1+cu117
• Datasets 2.9.1.dev0
• Tokenizers 0.13.2